A recent set of Siena/New York Times polls that showed an extremely tight presidential race in key battleground states has sparked discussions—and angst—across the internet. Matthew Ygelsias at Vox used these polls as a starting point to examine whether African American support for Donald Trump has increased. While Trump is of course deeply unpopular among black voters overall, his margins are what matter, as the Democrats’ path to victory in many battleground states next year will rely on running up the score with communities of color to offset losses among rural whites.
If in fact it were true that Trump has seen an uptick with black Americans, it would be an important factor in the race for the White House. The evidence, however, is not compelling. First of all, let’s tackle the idea that Trump performed better than Romney or McCain among African Americans. Yglesias cites the Washington Post’s Philip Bump, who last year pointed out that exit polling showed Hillary Clinton did 5 points worse among African Americans than Barack Obama did in 2012, while Trump scored 2 points better than Romney.
But exit polls have their flaws, and they’re not the only way to analyze the electorate. A report from Catalist, a Democratic data firm, examined a large voter database and concluded that there had in fact been no drop in Democratic performance with black voters: Clinton, it determined, had done as well as Obama, with both winning the group by 93 points.
Yet even if you feel that exit polls are preferable to voter files, Bump acknowledged that the issue for Clinton among African Americans was not that an appreciable number switched to Trump but rather than many who’d come out for Obama stayed home in 2016. What this means is that even if the exit polls are right and Clinton’s margins with black voters did shrink, that doesn’t necessarily mean Trump netted any additional votes. It goes without saying that African American turnout is a key issue for Democrats, but it’s very different from black voters actually warming to Trump.
Yglesias also points to a Catalist analysis of the 2018 elections in Minnesota and Georgia, which show Democrats doing slightly worse among African Americans than in 2016. In particular, the fact that Clinton outperformed Stacey Abrams in her race for governor seems shocking. But comparing presidential and midterm years can be treacherous. Midterm electorates are smaller, older, and invariably whiter than presidential electorates.
We can, however, see how 2018 compares to previous midterms. Digging deeper into Catalist’s nationwide data shows that Democrats overall won the African American vote by 90 points last year—16 points better than in 2014 and 10 points better than in 2010. Even accounting for shifts in the overall electorate (namely, that Democrats did much better in 2018 than 2014 or 2010), Democrats’ margin with black voters increased by 6 points from 2014 and decreased by only 2 points from 2010.
Finally, returning to the Times polls, they showed all three major Democratic candidates running behind Clinton’s 2016 margin among African Americans by 5 to 11 points. But the sub-sample of black voters was just 328 respondents across six different states. Compared to the entire sample, which included more than 10 times as many voters, the margin of error for such a small sub-sample is far higher.
By contrast, the Democratic pollster Civiqs, which has been polling Trump’s favorability rating continually for four years, found that on the eve of the 2016 election, 91% of black voters had an unfavorable view of Trump. Today, that figure stands at 92%. Similarly, a recent Fox News poll showed Trump getting just 6% of the vote in head-to-head matchups against both Elizabeth Warren and Joe Biden. To the degree that the 2020 Democratic candidates trail Clinton’s 2016 performance, it’s likely because lower-information voters are undecided rather than any sort of increase in Trump support.
It’s possible that Trump’s support among black voters has ticked up since 2016, but there’s not much of a case for it. And we shouldn’t rely on a single sub-sample of a single set of polls a year out before Election Day, when the evidence more broadly doesn’t back up this notion.